EHR/FHIR Patient Care and Alerts with GPT-4, Twilio, Gmail, and Slack

This n8n workflow automates the coordination of patient care and alerts by integrating with EHR/FHIR data, leveraging GPT-4 for intelligent processing, and dispatching notifications via Twilio (SMS), Gmail, and Slack. It's designed to streamline communication, ensure timely interventions, and enhance overall patient management.

What it does

This workflow is triggered manually and performs the following key steps:

1. Initiates Workflow: Starts the process with a manual trigger.
2. Fetches Patient Data: Connects to a PostgreSQL database to retrieve relevant patient information.
3. Processes with AI Agent (GPT-4): Uses an OpenAI Chat Model and a Structured Output Parser within an AI Agent to interpret patient data and generate actionable insights or alerts.
4. Filters Critical Alerts: An "If" node evaluates the AI Agent's output to determine if a critical alert condition is met.
5. Dispatches Notifications:
   • Critical Alerts: If a critical alert is identified:
     • Sends an SMS notification via Twilio.
     • Sends an email notification via Send Email (Gmail).
     • Posts a message to a designated Slack channel.
   • Non-Critical Information: If no critical alert is identified, the information is still processed and could be used for other purposes (though not explicitly routed to an action in the provided JSON).
6. Logs Activity: An HTTP Request node is included, potentially for logging the workflow's activity or interacting with an external system (though its specific configuration is not detailed in the JSON).
7. Aggregates Output: A Merge node and an Aggregate node are used to combine and structure the outputs from various branches of the workflow, preparing them for further processing or review.
8. Waits (Optional Delay): A "Wait" node is present, indicating a potential pause in the workflow, possibly for rate limiting or to allow time for external systems to process previous steps.
9. Sets Fields: An "Edit Fields (Set)" node is used to manipulate or add data fields within the workflow, likely preparing data for subsequent nodes.
10. Conditional Routing: A "Switch" node provides further conditional logic, allowing the workflow to branch based on specific data values.

Prerequisites/Requirements

To use this workflow, you will need:

• n8n Instance: A running instance of n8n.
• PostgreSQL Database: Access to a PostgreSQL database containing patient data.
• OpenAI API Key: For the AI Agent (GPT-4) node.
• Twilio Account: With configured phone numbers for sending SMS alerts.
• Gmail Account: Or an SMTP server configured for the Send Email node.
• Slack Account: With a workspace and channel for receiving alerts.
• Webhook Endpoint: If the "Webhook" node is intended as a trigger, you'll need an external system to send data to this endpoint.

Setup/Usage

1. Import the workflow: Download the provided JSON and import it into your n8n instance.
2. Configure Credentials:
   • Set up your PostgreSQL credentials to connect to your patient database.
   • Configure your OpenAI credentials for the AI Agent.
   • Set up your Twilio credentials (Account SID, Auth Token, and a Twilio phone number).
   • Configure your Gmail (or SMTP) credentials for sending emails.
   • Set up your Slack credentials and specify the target channel for messages.
3. Customize Nodes:
   • Postgres: Adjust the SQL query in the Postgres node to fetch the specific patient data required.
   • AI Agent: Configure the prompt for the OpenAI Chat Model to accurately interpret your patient data and define the expected structured output.
   • If Node: Define the conditions for critical alerts based on the output from the AI Agent.
   • Twilio, Send Email, Slack: Customize the message content and recipient details for each notification node.
   • HTTP Request: Configure the URL and payload if you intend to log data to an external service.
   • Edit Fields (Set): Adjust any data transformations as needed.
   • Switch Node: Define additional routing logic based on your specific requirements.
   • Wait Node: Adjust the delay duration if necessary.
4. Activate the Workflow: Once all credentials and configurations are set, activate the workflow.
5. Trigger the Workflow: Since this workflow is manually triggered, you will need to execute it manually within n8n. If a Webhook is intended as the primary trigger, ensure your external system is configured to send data to the Webhook URL provided by n8n.